Fact, Fiction, and Forecast: Fourth Edition

Nelson Goodman is not only an excellent philosopher, he's also an excellent writer. While the first chapter of this book may be difficult to slog through for those with little background in counterfactuals, the remaining three are clear, engaging, and insightful. (In fact, you can skip the first chapter and still profit immensely from the book.) Goodman begins with the classic problem of induction as outlined by Hume. He interprets Hume as having tried to solve the problem by describing the sort of induction we consider rational. He then uses this understanding as a springboard into a more complete theory of induction. While I have reservations about his interpretation of Hume, I find his approach to the problem of induction is fascinating.Before outlining his own proposal, Goodman raises the stakes by offering a new problem of induction. Goodman shows that induction must amount to more than a commitment to the past being like the future because, by defining new predicates in terms of the ones we typically use, we can argue on the basis of previous confirmation that just about anything will be the case. This is obviously undesirable, since we want induction to affirm that all ravens are black and not affirm that all ravens are white.Goodman's solution to this new problem is to define induction with reference not only to existing data, but also to past predictions (or more generally, projections, inductive claims about cases that have not yet been examined for confirmation). A given hypothesis or predicate becomes more "entrenched" by having been used previously in successful projections, giving the predicate or hypothesis greater legitimacy than the infinite other hypotheses that might fit the data, but which we have not previously affirmed.OF course, for the realist who believes we have predicates that cut the world up more or less as it is, the new problem of induction is no problem at all. But even still, Goodman's book is a fascinating exploration of a way of dealing with induction and a model of philosophical writing.

Due to this book's reputation and the fact that it is standard, contemporary reading material for philosophy students - I was expecting to be disappointed, that the book wouldn't live up to its prestige. I was wrong. Incredibly wrong. "Fact, Fiction, and Forecast" lives up to Hilary Putnam's description as a "contemporary classic."Goodman reformulates the problem of induction using his conceptual ideas "grue" and "bleen." These ideas are time dependent and so require a date to keep in mind - let's say January 1st of the year 2017. An object is "grue" if it is observed on or before January 1st 2017 and is green or after our chosen date and is blue. Similarly, an object is "bleen" if it is blue on or before January 1st 2017 or green after the same date.The consequence is that after our given date, objects which were previously described as "green and grue" such as our dear friend the US dollar are now "green and bleen" while objects which were previously known as "blue and bleen" are now "blue and grue."Goodman's problem can be stated as such - how can we know which set of predicates is best projectable into the future: blue and green or grue and bleen?Goodman's final chapter "Prospects for a Theory of Projection" attempts to address this problem - as a fellow reviewer on Amazon put it: "It is in this chapter that we see the full philosophic mind of Goodman. The depth and relentless thought that Goodman puts into this chapter will forever 'entrench' his name in the philosophic discipline."

Goodman starts out with an attempt to tackle the problem of interpreting counterfactuals ("What would happen if X would be the case instead of Y."). He doesn't solve the problem but gives some deep insights, especially on the connection between counterfactuals and scientific laws. In the next section he tries to tackle the problem by taking a look at a specific sort of counterfactuals, dispositional predicates. These are predicates like "flexibility" ("If I would bend this..."). He does tackle that problem. He doesn't use strange concepts like "possible worlds", that are more problematic than the original problem, but shows how dispositional predicates can be interpreted as statements about past observations, which reduces the problem to the good old problem of induction, which he adresses in the third section. He argues that Hume has solved the problem on how we can know that the future will behave like the past (we simply can't). The real question is not justifying induction but describing how it is done. Several people have attempted to do just that and Goodman discusses their work in some detail. He shows that there is a new, deeper problem: How can we separate theories about predicates ("All X are Y.") from these predicates. He constructs a strange predicate, grue, that is green until some future time t and blue afterwards. The theory "All emeralds are green." is as well supported as the theory "All emeralds are green." One can also construct "blue" and "green" from "bleen" and "grue", so the choice of predicates seems to be somewhat arbitrary. It is easy to construct similar predicates and noone has found a general way to rule them out yet. So how can we decide what predicates we should use in our theories? Goodman argues that this is pure convention, based on tradition. Not everyone will accept this answer (I don't), but this isn't necessary for seeing the brilliance of this work.

En este librito, Goodman aborda, en primer lugar, el problema de los condicionales contrafactuales (“si hubieras frotado ese cerillo entonces habría prendido”); así como la proyección futura de ciertos enunciados y la validez que solemos asociar con los mismos (“el nuevo enigma de la inducción” es un clásico).De lectura amena, sin grandes requerimientos técnicos en lógica simbólica y (casi) completamente autocontenido, este libro —en palabras del presentador, el filósofo Hilary Putnam—, “es una lectura obligada para cualquier estudiante serio de Filosofía en nuestros días”. Desde mi punto de vista tiene razón.

まぎれない名著です。心の哲学,人工知能,脳理論に重要な示唆を提供しています。非常によい日本語訳もあります。脳は"事実(経験的事実)"を知覚する。脳は、"虚構(仮定)"を作り出したり、未来の"予言(予測)"を作り出したりしない。"虚構(仮定)"を作り出したり、未来の"予言(予測)"を作り出したりしているように観える主張も、"事実(経験的事実)"を述べている。例えば、「このゴミは可燃物だ!」という"仮定"に観える主張は、そのゴミが 燃える材質であるという"事実(経験的事実)"を述べている。空を見上げての「午後は 雨が降るだろう!」という"予言"に観える主張は、現在の空が 曇っていて暗いという"事実(経験的事実)"を述べている。脳は、"事実(経験的事実)"を、ある行動を誘発するディスポジション(アフォーダンス)に結びつけている。脳が持つ"知識"とは 正に これである。グッドマンは、これが "科学的知識"のエッセンス(本質的要素)であると言う。この本でグッドマンがとっている態度(姿勢)は極めて納得が行くものである。人工知能や脳の機能に関心を持つ人に是非お薦めする。※脳は、メンテナンスが不可能な世界モデルを持たない。※[補足] 今日のディープ・ラーニングの成功は、それが 正に "事実の知覚"を実現している故である。問題は、「行動に向けて どのような事実(特徴量)を知覚すればよいか？」である。

bargain